[Objective] To understand the surface moisture conditions and to implement precise field irrigation, spectral characteristics of soil moisture in salinized soil and model fitting accuracy were analyzed in the Northern Yinchuan City of Ningxia Hui Autonomous Region. [Methods] With the severe salinized soil in the Northern Yinchuan City as the subject, a variety of mathematical transformations were carried out on the raw spectral reflectance of soil moisture. The stepwise regression (SR) and the grey correlation degree (GCD) were used to screen sensitive wave bands, and then the multiple linear regression (MLR), partial least-squares regression (PLSR) and support vector machine (SVM) were used to calculate the fitting accuracy model of soil moisture content (SMC). [Results] ① The soil spectral reflectance decreased with the increase of SMC when SMC was below 26.34%, and soil spectral reflectance increased with the increase of SMC when SMC was higher than 26.34%. The change of reflectance in the NIR region were larger than that in visible region, and the spectral characteristic curves showed obvious absorption bands at 1 460 nm and 1 950 nm when continuum removed (CR) was used. ② Different transformation methods of the reflectance had different fitting accuracy about MLR, PLSR and SVM models, the overall fitting capacity of SVM model was better than MLR and PLSR models. Except for the GCD-SVM model by the reciprocal reflectance (RR) transformation, the R_C² and R_P² of the SVM models range from 0.943 7 to 0.999 5 and have high fitting accuracy. ③ In the SVM models, the GCD-SVM model based on first derivative of logarithmic reflectance (FLR) transformation had the highest determination coefficient (R_C² was 0.987 4 and R_P² was 0.999 5), which was the best fitting model of SMC for severe salinized soil. [Conclusion] The SVM model was the best model for SMC, it could accurately predict the surface moisture in severe salinized soil in Northern Yinchuan City of Ningxia region.